ImmGen report: sexual dimorphism in the immune system transcriptome

Sexual dimorphism in the mammalian immune system is manifested as more frequent and severe infectious diseases in males and, on the other hand, higher rates of autoimmune disease in females, yet insights underlying those differences are still lacking. Here we characterize sex differences in the immune system by RNA and ATAC sequence profiling of untreated and interferon-induced immune cell types in male and female mice. We detect very few differentially expressed genes between male and female immune cells except in macrophages from three different tissues. Accordingly, very few genomic regions display differences in accessibility between sexes. Transcriptional sexual dimorphism in macrophages is mediated by genes of innate immune pathways, and increases after interferon stimulation. Thus, the stronger immune response of females may be due to more activated innate immune pathways prior to pathogen invasion.

5. The methods in this manuscript are poorly described. With this impressive number of samples and that IMMGEN aims to provide a standard for the field, it is essential to describe the experimental setup that went into each analysis: (a) The RNA-seq experiment design is not described clearly in the text, figure 1A, or the methods. For example, Figure 1A shows the clustering for 11 different cell types. There is a lot of information that is unknown such as: (1) Per sample, how many replicates are female? (2) How many replicates are male?
(3) Are the RNA-seq samples derived from pooled mice? If so, how mice many per pool? (4) The methods section indicates that "young," "adult," and "aged" mice were used. How many replicates are from each age group?
(b) In the pan-immune approach, that identified 14 SDEGs, it is unclear what comparison was used in order to identify this group of 14 genes. This is in part because from Figure 1A and from the methods section, it is also unclear what exactly comprises dataset A and B and how they differ. "The panimmune approach was implemented by applying a paired t-test to the RNA sequencing datasets described above (datasets A and B, Fig. 1a), in which samples were matched by cell type, age and dataset." (c) The flow cytometry methods section is missing. What surface markers were used to sort each cell population?
(d) The methods describing the in vivo INFalpha experiment are unclear and missing detail. This comment specifically refers to the part: "we analyzed RNA-seq profiles of three cell types, GN, MF and B cells, from male and female mice before and after exposure to IFNα." From the methods section, which describes: "For stimulation, mouse were intraperitoneal injected with 1k and 10k Interferon alpha corresponding to 1,000 and 10,000 enzyme units and sacrificed two hours later." What is 1k and 10k referring too? What was the manufacture of INFalpha? How many mice were injected? Males, females, age? The authors must include this information in a clear manner.
(e) The authors must clarify what strain of C57BL6 was used J or N (or a mix) since there are reported differences in blood cells between J and N strain and because IMMGEN is a standard for the field. The methods only lists "C57BL6." Elevated Oxidative Stress Impairs Hematopoietic Progenitor Function in C57BL/6 Substrains PMID: 30017822 6. This paper involves mouse experiments, but there is no mouse ethic statement, no statement on what type of facility the mice were housed in (pathogen-free, pathogen-specific, etc), and there is no institutional oversight statement from IACUC in the methods section. These statements must be included.
7. The authors should put their findings into the context of what is known about sexual dimorphic gene expression in mammals. This area is of interest because there appears to be conflicting studies --some show mammals have dramatic differences in gene expression between males and females and others do not. This study is one of the more complete and modern and of the opinion that this finding alone is very important. Here are other studies that started to debate "how sexually dimorphic" geneexpression: Sexual dimorphism in mammalian gene expression. PMID: 15851067 Major molecular differences between mammalian sexes are involved in drug metabolism and renal function. PMID: 15177028 Sex-Specific Selection and Sex-Biased Gene Expression in Humans and Flies https://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.1006170 Tissue-specific expression and regulation of sexually dimorphic genes in mice PMID: 16825664 8. In this study, they re-analyze existing ATAC-seq data (IMMGEM) and point out that there are some differences in open chromatin regions between males and females -Firre, GM35612, and Pls3 (With respect to Fig.5C,D). The authors should put this finding into more of a biological context. For example (1) previous reports find that Firre excapes X inactivation and there are regulatory differences in the inactive and active X. This was determined by allele specific chip and RNA seq. Function and evolution of local repeats in the Firre locus PMID: 27009974 Topological organization of multichromosomal regions by the long intergenic noncoding RNA Firre PMID: 24463464 Regulation of the ESC transcriptome by nuclear long noncoding RNAs PMID: 26048247 The lncRNA Firre anchors the inactive X chromosome to the nucleolus by binding CTCF and maintains H3K27me3 methylation PMID: 25887447 MINOR POINTS: 1. What do the 5 different modules represent? It is not clearly stated. "The highly complicated interferon response of the same cell types as here was recently parsed into five modules, that are different in terms of the cells in which they are active, their regulation, their tonic interferon response, and more." 2. Missing Citation: "An example of this potential effect is the X-linked gene DEAD-box helicase 3 (Ddx3x) and its Y-linked homolog (Ddx3y). Ddx3x is crucial for interferon (IFN) production in response to pathogens, and high levels of Ddx3x can boost the female IFN inducers response." Reviewer #2 (Immune ageing, systems immunology)(Remarks to the Author): The manuscript addresses the highly relevant question of sexual dimorphism in immune responses and autoimmune. The approach taken to extend the transcriptome data base of the Immunological Genome Project to include female mice and extend this study to include ATAC-seq data and transcriptional data after in vivo challenge is timely. Data generation followed standard operating procedure developed by ImmunGen and quality control and analytical tools appear to be overall appropriate. The most striking finding is that gender differences were small, of borderline significance and not consistent across tissue, even not related myeloid tissues. 1. Given the small differences, a discussion on the power of the study would be desirable. 2. Given that gender signatures are tissue specific, the potential limitation that cells were not stratified for differentiation status should be discussed. For example, there are substantial differences between naïve and memory T cells in transcriptomes and chromatin accessibility maps. 3. The study included three different age groups, but age is only mentioned in passing in the result section. If age is important variable, could the inclusion of three different age groups reduced the powered? 4. What was the rationale for using different FDRs for different comparisons? 5. Although of interest, the comparison to SLE transcriptome analysis remains superficial. It is unclear whether the two chosen comparison datasets are representative. There is no discussion on the heterogeneity of cell types in PBMC that account for the SLE signature in the chosen datasets and the SLE score is not well explained. 6. The findings on Firre are of interest, but preliminary and should not be overinterpreted. It is certainly not justified to make Firre a focus point in the abstract. 7. The data appear to be consistent with the interpretation that female cells are less able to maintain quiescence and macrophages are the more responsive cell type. Are the observed signatures truly an IFN response signature or do they just reflect macrophage activation? 8. The authors should discuss that the human studies but not the mouse studies show a signature for T cells. The manuscript by Howard Chung's group in Cell Systems that ATAC-seq signatures in PBMCs are driven by gender should be discussed. Minor comments Incomplete sentence in the introduction (Landscape of X chromosome inactivation across human tissues).

Reviewers' comments: Reviewer #1 (RNAseq, systems biology)(Remarks to the Author):
SUMMARY It is well-established that females and males have different immune responses, yet studies addressing the underlying causes at scale are lacking. Gal-Oz et al., set out to determine the differences between male and female immune cell types and take a gene-expression based approach. Surprisingly, they find that across 11 different cell populations, only one cell type, MFs, displays sex-specific differences in gene expression. The 41 gene cohort they identify in MFs shows a female-specific enrichment in IFN-stimulated genes. And the authors show that these genes are found at higher basal level expression, and suggest that these findings lead to female immune alertness. The central finding in this paper that there is a group of sexspecific genes in MFs that could help explain immune differences in males and females. This is a comprehensive study that provides a key insight that, as measured by RNA, the immune system is not dramatically sexually dimorphic with the exception of Macrophages. This study will be of great interest of the Nature communications readership. Below are the following considerations for the authors. We acknowledge this point should be clarified, and clarified accordingly in the results section 'Cell-type-specific sex signature' (p.8). In both pan immune and cell type specific tests, we paired the samples by age, as now clearly stated in text. We added a new supplementary figure 2 of the SDEGs heatmaps with colorbars indicating sex, age and dataset to address this point (Suppl. Figure 2A). As Figure  1B below shows, the standard deviation of the 41 SDEGs in the three samples that are of the same age is mostly similar to that in the three samples from different ages for males. For females (Figure 1A below), the old female is a bit of an outlier (as can be seen in paper Figure 3b and Suppl. Figure 2A) -not enough to be removed, but enough to drive the standard deviation of 5 SDEGs to be higher in females. We also show below, two figures ( Figure 1C-D) with the patterns of all 41 SDEGs (separately for male and female SDEGs). Both show that the 41 SDEGs are differentially expressed between male and female, independent of age.
However, as the reviewers predicted, some genes seem to be more different in specific ages. Unfortunately, our design cannot robustly identify such effects, but we hope to further study them in future studies. We thank the reviewer for these excellent and relevant references, and added them to the introduction (P.5 ) and discussion (P.15-16).
Regarding TLR4, the protein expression is indeed higher in resting male macrophages compared to females, (as reviewed in PMID: 28775365), but there is no difference at the mRNA level, both in the original paper (PMID: 16574244) and in our data. We added this in the discussion, as one possible explanation to the scarcity of the differences we see.

We added the suggested information as a Supplementary Note (Supplementary Note 2), by a comprehensive search of immune related phenotypes of each gene in
PubMed, OMIM, gene cards and JAX. We have also added a discussion about two FCGRs that are expressed higher in females (discussion P.16).

The methods in this manuscript are poorly described. With this impressive number of samples and that IMMGEN aims to provide a standard for the field, it is essential to describe the experimental setup that went into each analysis: (a)
The RNA-seq experiment design is not described clearly in the text, figure 1A, or the methods. For example, Figure 1A shows the clustering for 11 different cell types. There is a lot of information that is unknown such as: (1) Per sample, how many replicates are female? (2) How many replicates are male? (3) Are the RNA-seq samples derived from pooled mice? If so, how mice many per pool? (4) The methods section indicates that "young," "adult," and "aged" mice were used. How many replicates are from each age group?

(b)
In the pan-immune approach, that identified 14 SDEGs, it is unclear what comparison was used in order to identify this group of 14 genes. This is in part because from Figure 1A A and B, Fig. 1a), in which samples were matched by cell type, age and dataset."

and from the methods section, it is also unclear what exactly comprises dataset A and B and how they differ. "The pan-immune approach was implemented by applying a paired t-test to the RNA sequencing datasets described above (datasets
We clarified the text result section-Pan-immune sex signature, p.6.
" Datasets A and B were produced independently, and are different in the ages of the samples, the protocols and the depth of sequencing. The pairing was done by dataset and age, to account for batch effects and age related changes, respectively. Though age related changes could be sex dependent, the number of samples in each age group does not allow identification of such effect accurately."

(c)
The flow cytometry methods section is missing. What surface markers were used to sort each cell population?
We apologize for omitting this information. The surface markers used for each population have been added to Supplementary table 1, as well as the antibodies used. We also added relevant references to the table in method section: "RNA sequencing data generation".

Data was generated according to the ImmGen SOP. To clarify this point we added:
"Mice were sacrificed and immunocytes were isolated to high purity by flow cytometry according to the ImmGen SOP ( https://www.immgen.org/ ). For stimulated and unstimulated mice, spleen, whole CNS and peritoneal cavity were harvested." The methods describing the in vivo INFalpha experiment are unclear and missing detail. This comment specifically refers to the part: "we analyzed RNA-seq profiles of three cell types, GN, MF and B cells, from male and female mice before and after exposure to IFNα." From the methods section, which describes: "For stimulation, mouse were intraperitoneal injected with 1k and 10k Interferon alpha corresponding to 1,000 and 10,000 enzyme units and sacrificed two hours later." What is 1k and 10k referring too? What was the manufacture of INFalpha? How many mice were injected? Males, females, age? The authors must include this information in a clear manner.
We apologize for omitting this information. IFNa manufacturer is R&D SYSTEMS Catalog #12100-1. We have organized the section and added the missing information to the Methods section "RNA sequencing data". Table  1). For stimulation, 12 mice were intraperitoneal injected with IFN-alpha (R&D SYSTEMS Catalog #12100-1), 3 males and 3 females were injected with 1,000 (1k) or 10,000 (10k) enzyme units, and mice were sacrificed two hours later. The 24 stimulated samples comprised of male and female triplicates for 1k IFN (B cells only) and 10k IFN (B, GN and MF)."

(e) The authors must clarify what strain of C57BL6 was used J or N (or a mix) since there are reported differences in blood cells between J and N strain and because IMMGEN is a standard for the field. The methods only lists "C57BL6."
We apologize for omitting this information. We added a "Mice" paragraph to the method section, which includes the mice strain. "C56BI/6 J inbred mice were obtained from the ImmGen Colony at Jackson Laboratory, Maine"

Elevated Oxidative Stress Impairs Hematopoietic Progenitor Function in C57BL/6 Substrains PMID: 30017822
We thank the reviewer for pointing out the difference between the B6 substrains, and add the substrains difference and the suggested reference (as well as two others showing stroke outcome and neurological performance differences between strains) and another one to the discussion on page 18.
6. This paper involves mouse experiments, but there is no mouse ethic statement, no statement on what type of facility the mice were housed in (pathogen-free, pathogen-specific, etc), and there is no institutional oversight statement from IACUC in the methods section. These statements must be included.
We apologize for omitting this information. In the newly added "Mice" paragraph we added housing and IACUC protocol. " C56BI/6 J inbred mice were obtained from the ImmGen Colony at Jackson Laboratory, Maine, and housed in a full barrier facility. Ethics oversight by Harvard Medical School, under Institutional Animal Care and Use Committee protocol IS1257." We added those publications to the discussion section P.17. We also updated the reference to the now published ImmGen ATAC-seq paper.

The authors should put their findings into the context of what is known about sexual dimorphic gene expression in mammals. This area is of interest because there appears to be conflicting studies --some show mammals have dramatic differences in gene expression between males and females and others do not. This
MINOR POINTS: 1. What do the 5 different modules represent? It is not clearly stated. "The highly complicated interferon response of the same cell types as here was recently parse into five modules, that are different in terms of the cells in which they are active, their regulation, their tonic interferon response, and more." We only use the modules to put the interferon response here in a wider context of the general interferon response. We have now edited the description of the modules creation (P.9) to add more details from the original study defining the modules (Mostafvi et al), namely: "Mostafavi et al. built a regulatory network between 92 predicted regulators and 102 interferon stimulated genes, with 2,691 links connecting regulators to genes based mostly on co-expression in human and mouse response to interferon in multiple datasets. The 102 interferon stimulated genes in the regulatory network were then parsed into five distinct modules based on similarity of their regulators -C1-C5. Thus, the modules are different in terms of regulation by definition, but turned out to be informative in that they are also distinct in the cell types in which they are active, their tonic interferon response, and more. For example, STAT1/2 and IRF9 were the main predicted regulators C3 and C4, and accordingly those modules were enriched for core ISRE motif. Module C3 contained mostly antiviral effectors and key positive and negative regulators. C1 and C2 were enriched for RNA processing, C4 for metabolic regulation, and C5 for inflammation mediators or regulators." 2. Missing Citation: "An example of this potential effect is the X-linked gene DEADbox helicase 3 (Ddx3x) and its Y-linked homolog (Ddx3y). Ddx3x is crucial for interferon (IFN) production in response to pathogens, and high levels of Ddx3x can boost the female IFN inducers response." We rephrased the sentence to clarify which information comes from which source, and added a recent publication showing sexual dimorphism in the response to infection in DDX3Y hematopoietic KO (PMID: 30475900) " Another example of this potential effect is the X-linked gene DEAD-box helicase 3 (Ddx3x) and its Y-linked homolog (Ddx3y). Ddx3x is crucial for interferon (IFN) production in response to pathogens 17 and in high levels can boost the female IFN inducers response. Indeed, mice lacking Ddx3x in hematopoietic cells have higher susceptibility to Listeria monocytogenes and reduced numbers of lymphocytes, not compensated by Ddx3y 18 "

Reviewer #2 (Immune ageing, systems immunology)(Remarks to the Author):
The manuscript addresses the highly relevant question of sexual dimorphism in immune responses and autoimmune. The approach taken to extend the transcriptome data base of the Immunological Genome Project to include female mice and extend this study to include ATAC-seq data and transcriptional data after in vivo challenge is timely. Data generation followed standard operating procedure developed by ImmunGen and quality control and analytical tools appear to be overall appropriate. The most striking finding is that gender differences were small, of borderline significance and not consistent across tissue, even not related myeloid tissues.
1. Given the small differences, a discussion on the power of the study would be desirable.
We agree -given our small sample size, especially for the cell type specific effect, we can only identify relatively large effect sizes. In the initial submission we were satisfied with that, as the known sex related genes (XIST on chr x, Eif2S3Y on chromosome Y) were clearly identified in both pan-immune analysis and cell type specific analysis. Additional differentially expressed genes were identified in the pan immune analysis, and for macrophages. As macrophages had the same testparameters as the other cell types, the fact that for the other cell types no SDEGs were identified is not solely due to lack of power. As also seen in the human analysis of the much larger ImmVar data, the magnitude of differences between sexes is apparently small, and requires larger sample size.
Following the reviewer comment, we learned that a formal power analysis for a multiple test setup of thousands of tests with FDR is non trivial. The power to detect a specific effect in a given test depends on the results of all tests. In our case, it means that a gene with a given t-statistic in the per-cell type setting may be found significant at one cell type and not significant in another with the same FDR threshold.